Wind power generation prediction utilizing Kolmogorov-Arnold networks
The increasing integration of wind energy into power grids necessitates accurate and reliable wind power generation forecasts. Precise predictions are crucial for grid stability, efficient energy management, and optimal economic dispatch. However, the inherent variability and intermittency of wind r...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Institute of Physics
2025
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/44974/ http://umpir.ump.edu.my/id/eprint/44974/1/Sulaiman_2025_Eng._Res._Express_7_025335.pdf |
| Summary: | The increasing integration of wind energy into power grids necessitates accurate and reliable wind power generation forecasts. Precise predictions are crucial for grid stability, efficient energy management, and optimal economic dispatch. However, the inherent variability and intermittency of wind resources pose significant challenges to traditional forecasting methods. Recently, Kolmogorov-Arnold Networks (KAN) have gained attention in the machine learning community for their ability to model complex, non-linear systems. This study explores the use of KAN for wind power generation prediction, utilizing its distinctive architecture to model complex patterns within wind power data. The performance of KAN is compared against established machine learning approaches, namely Neural Networks (NN) and Long Short-Term Memory (LSTM) networks, using a comprehensive dataset of weather and turbine parameters collected over two years. The results demonstrate that KAN exhibits superior performance in terms of prediction accuracy and consistency. Specifically, KAN achieved the best RMSE of 87.5, MAE of 61.4, and an R2 value of 0.9723, indicating high accuracy and reliability. KAN�s narrower distribution of residuals centered closer to zero compared to LSTM indicates more reliable predictions. While NN displayed the sharpest peak in error distribution, suggesting high consistency for certain ranges, KAN provided a better balance between accuracy and adaptability across various prediction scenarios. These findings suggest that KAN offers a promising approach for wind power forecasting, potentially improving grid integration strategies and operational efficiency in wind energy systems. |
|---|